AI Paper Atlas

Research map / Architectures and efficiency / Neural architecture search

EC-DARTS: Inducing Equalized and Consistent Optimization Into DARTS

Qinqin Zhou, Xiawu Zheng, Liujuan Cao, Bineng Zhong, Teng Xi, Gang Zhang, Errui Ding, Mingliang Xu, Rongrong Ji

ICCV 2021 · Architectures and efficiency › Neural architecture search · approach: Bi-level differentiable NAS · 9 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add Cross-Edge Normalization (CEN) (ablation)DARTS · CIFAR10test error (%): 2.76 -> 2.60Improved
add Induced Structural Transition (IST) (ablation)DARTS · CIFAR10test error (%): 2.76 -> 2.47Improved
add IST on top of CEN (ablation)DARTS + CEN · CIFAR10test error (%): 2.60 -> 2.39Improved
add CEN on top of IST (ablation)DARTS + IST · CIFAR10test error (%): 2.47 -> 2.39Improved
add CEN (ablation)DARTS · CIFAR10 architectures (10 edge combinations)Kendall Tau (operation weight vs performance): 0.06 -> 0.63Improved
add IST (ablation)DARTS · single search, 10 sampled architecturesKendall Tau (search rank vs retraining rank): 0.13 -> 0.47Improved
use EC-DARTS instead of DARTSDARTS · 10 independent searches on CIFAR10average retraining accuracy (%): 97.13% -> 97.61%Improved
add CEN (ablation)DARTS · CIFAR10search costNo clear effect
search on target dataset instead of transferring from CIFAR10EC-DARTS · Tiny-ImageNet-200test error (%): 28.7 -> 28.1Improved
search on target dataset instead of transferring from CIFAR10EC-DARTS · ImageNettop-1 error (%): 24.3 -> 23.6Improved
search on CIFAR100 instead of transferring from CIFAR10EC-DARTS · CIFAR100test error (%): 16.13 -> 15.72Improved
search on CIFAR100 instead of CIFAR10EC-DARTS · CIFAR10test error (%): 2.39 -> 2.44Mixed

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy